Project Summary: Software as a Service (SaaS)
Within MVM College, Research Computing has often focused on data storage (DataStore) and data processing (EDDIE). However, there is also a smaller niche requirement to host applications. Some of these are commercial products, others have been developed in-house by University IT professionals and Research Software Engineers. To date, we have often turned to the excellent VMWare Server Virtualisation and Hosting Service, details of which can be found here:
However, within the Research Services section of Information Services, we have a new private cloud server hosting service. Details of the 'Eleanor' service can be found here:
The goal of this project was to take two exemplars of 'Software as a Service' and to support their adoption of the 'Eleanor' service.
R Studio Pro Server
The first exemplar was the desire to create a "R Studio Server Pro" instance so that Students could be taught how to programme in R using a web based user interface. There was a need to have R available to undergraduates and post-graduates within the University. At that time, there was no version of R available on the computer labs nor staff desktops. The web based version also had the advantage that Students could connect from anywhere using just about any web browser to continue their R studies.
Voice Banking Project
The Voice Bank ing Project is a research project. More information about their project can be found here:
In its simplest forms, this project helps by digitally capturing a patient voice before they loose their voice to neurological diseases and deterioration. The captured voice can then be repaired with sounds from other donors and a new synthetic voice can be created using the massive processing power of the Eleanor cloud machines. This synthetic voice can then be used on speech assistants / voice synthesiser devices. Unlike the technology of the past, this approach provides the patient with a voice that sounds like them; which has been identified as an important for their wellbeing and preserving their sense of identity for their family and friends.
Results & Findings
R Studio Pro Server
There are more notes available for this project, they can be found under the Other Resources link.
Initially this was built as a very powerful single machine with lots of cores and lots of memory. However, this didn't quite provide the expected performance due to the underlying R Studio Pro system only running as a single threaded application. The choice was then made to build it as a cluster of small machines; the master device would receive requests to use R Studio and it would then federate the request to a worker node / device. This provided excellent scaling through clustering technologies. Essentially embracing the nature of cloud computing.
This system has supported the needs of Student training in R programming well.
(It should be noted that the RStudio desktop IDE is now available within the Supported Desktop application store)
The Voice Banking server has many computer services all incorporated into a single 'big' machine.
The system comprises of:-
1. A web server to host an application to help manage the Donors and Patients visits to the 'recording studios' to capture their voices.
2. A database server to hold meta-data.
3. A file server to securely store the voice recordings. This includes a file upload interface to receive the voice recordings from a bespoke iPad application.
4. A compute / processing platform to dynamically generate the synthetic voices.
This system has helped underpin the Voice Banks clinical and research ambitions to scale up their operations of capturing, banking and generating new synthetic voices for patients.
Despite initially appearing to be rather similar to the Enterprise VMWare service, Eleanor (underpinned by OpenStack) does offer a responsive and flexible platform. Everyone found it very easy to add capacity and performance to their cloud machines without much involvement of the service providers.
Both exemplars delivered tangible benefits within a short space of time.
There has been virtually zero downtime for this service during our engagement.
The support and availability of the Research Services team to assist has been perfect; many thanks to this team, especially for Jan Winters!
Elearnor has gone from strength to strength and we can highly support its further adoption accross the University.
Analysis of Resource Usage:
Staff Usage Estimate: 20 days
Staff Usage Actual: 30 days
Staff Usage Variance: 0%
Other Resource Estimate: 3 days
Other Resource Actual: 2 days
Other Resource Variance: 0%
Explanation for variance:
Key Learning Points:
When hosting applications, it is important to understand how they have been developed and if they would benefit from large multi-core virtualisation systems.